A learning record that follows the learner into your product.
Seneca is a learning record the learner controls. It keeps what each product observed, where it came from and how certain it is. When a learner arrives in your product and gives permission, you read a short, sourced picture of them and decide what to teach from there.
Learning evidence usually stays inside the product that collected it.
A score or grade rarely tells the next product what the learner saw, what help they had, how the question was built or how much uncertainty sits behind the result. So the next product runs its own diagnostic, and the learner repeats work they have already done.
Scores are conclusions. Seneca keeps the evidence underneath them, so the next product can check it for itself.
What goes in: the events your product already records.
You send what happened, as your product sees it today. There is no new test to build and no universal score to adopt.
Answers
Each response and its result: correct, incorrect or unscored, plus retries and the attempt number.
Timing
When the learner answered and how long each question took.
Assistance
The help that was present at the time, such as a hint, a teacher or an AI tool.
Item context
What kind of question it was, what it showed the learner and your own label for it. Your taxonomy stays intact.
Help requests
Hints the learner asked for, worked examples they opened and tutor chats.
Send them in the format you already have. A CSV or JSON export works.
If your product already produces xAPI or Caliper statements, those work too. You don't need to adopt a standard to start.
What comes back to your product.
With the learner's permission, Seneca returns a short picture your team or your tutor model can read in a few seconds, with everything behind it attached.
A short learner summary
Three or four lines a teacher or a tutor model can read.
Supporting evidence
The events behind each line, with counts.
Confidence and uncertainty
Strong conclusions are marked apart from the ones that still need checking.
Relevant prior interventions
What helped before, and where it was seen.
A suggested first step
A starting point your product can take, adapt or ignore.
Provenance for every conclusion
Which product it came from, when it was observed and how many observations support it.
Seneca provides a starting point. Your product decides what to teach.
One student, two apps, one evening.
This example comes from the interactive demo. The student and both apps are fictional.
An assessment app learns something
At school, the student gets 8 of 10 fraction questions right and 2 of 5 word problems. The last three word problems were answered unusually fast, 6 to 9 seconds each. A worked example helped twice.
Seneca keeps it with its source
Fraction procedures appear secure, with high confidence. Word problems need checking. Worked examples are noted as something that helped. Each conclusion carries the assessment app, the time it was observed and how many responses support it.
A tutoring app starts from it
That evening the student allows a tutoring app to read the math context. It opens with one short word-problem check. Without Seneca, it starts at question 1 of 20 and reaches word problems at question 14.
47 seconds, captions only. A fictional student and fictional apps.
Click through it yourself.
In the demo you can send the assessment results, open the evidence behind each conclusion, allow or deny the tutoring app's request, and switch Seneca off and on to compare the tutoring app's first minute.
About two minutes to click through. The starter kit shows the same idea with your own export format, using made-up data.
For learners, the record is free and theirs to control.
A learner's evidence moves only when the learner allows it. Each product asks for what it needs, and the learner can see and change that at any time.
Free to carry
Learners pay nothing to hold their record or take it from one product to the next.
Export
A learner can export their record.
Grant and revoke
Each product asks for access separately. The learner allows or denies it, and can revoke it later.
Private stays private
Things like a reading journal stay locked unless the learner chooses to share them. A tutoring app that asks for math context gets math context and nothing else.
What travels, and what stays with you.
Seneca standardizes what is known and leaves the next move to your product, your model and your teachers.
What can travel
- Outcome, timing and attempt history
- Assistance state and testing conditions
- Observable item and task features
- Source, scope and uncertainty
What stays with you
- Your curriculum and teaching sequence
- Your interface and recommendation logic
- Your proprietary questions and content
- Anything the learner did not authorize
Learners carry Seneca free. Companies pay to use it.
A portable record only works if no single vendor can trap the learner inside it. Companies pay for the infrastructure they need to read and write that record in production.
Learners
Holding the record, exporting it and revoking a product's access stay free.
Companies
Hosted reads and writes, permissions, connectors, audit, compliance and support.
The learner’s record
Learners can see, export and revoke their record at any time, and exports use open standards such as xAPI. Partners get the integration contract in a sandbox, under agreement.
Test Seneca with a fictional learner.
Send one made-up student history shaped like your data. We return a learner summary, the evidence behind it and an example first step inside your product. Fake data only, no production access, no engineering sprint, no commitment to a full integration.
Questions first? Write to kevinchoi@vindicaseneca.com.